UlazAI developer docs

Video errors and limits

Validation and integration limits

Enforce client-side validation before sending requests. Most failures come from invalid model combinations (duration, ratio, quality, or input type).

Common HTTP responses

Code Meaning Typical cause
400 Invalid request Unsupported ratio/duration/quality, missing prompt, invalid URL
401 Unauthorized Missing session or API auth context
402 Payment required Insufficient credits for requested job batch
404 Not found Unknown model slug or unknown job ID
429 Rate limited Per-user request limits exceeded

Model-specific guards

  • model_slug must exist in the video model matrix.
  • aspect_ratio and duration_seconds must match model options exactly.
  • quality_mode is required when a model has quality presets.
  • Hailuo 2.3 requires one input_image_urls entry and does not support two images.
  • wan_2_6_v2v requires one input_video_urls entry.
  • Kling 3.0 supports optional start_frame_url, end_frame_url, and elements with strict shape validation.
  • Batch size is capped to 4 jobs per request.

Prompt length limits

  • Kling 3.0 and Kling 2.6: 2500 characters.
  • Grok Imagine Video: 5000 characters.
  • Hailuo 2.3 Standard/Pro: 5000 characters.
  • Most remaining models follow standard prompt validation in the generation services.

Rate limits

  • /api/v1/video-studio/generate/: 24 POST requests per user per hour.
  • Video Studio tool endpoints: 24 POST requests per user per hour.
  • Veo 3.1 extend and upscale actions: 48 POST requests per user per hour.

Retry strategy

  • Do not retry 400 or 404 without fixing request data.
  • Do not retry 402 until the account has enough credits.
  • Retry 429 with backoff and respect server rate limits.
  • For transient failures, retry only after checking current job status to avoid duplicates.